Methodology & content principles

1. 01

   Findings carry evidence

   A report should show what was observed, where, and under which rule — not just a verdict. If we cannot show the evidence, we do not publish the claim.

2. 02

   Three kinds of statements, kept apart

   Rule-based checks are machine-verifiable. Content suggestions need human judgment. AI answer observations are samples with conditions. Mixing them into one score hides what each can tell you.

3. 03

   Numbers come with denominators

   “Mentioned in 6 of 18 valid responses, 2 runs failed and were excluded” is the shape of an honest metric. Rates without sample sizes and failure handling are not published here.

4. 04

   No guarantees

   We do not promise rankings, AI recommendations, or traffic. Anyone who guarantees AI visibility is selling certainty that does not exist.

5. 05

   Dates reflect reality

   Guide update times come from the actual content history in Git, not from the build date.

6. 06

   Sample data stays labeled

   Demos and exports use fixed sample data on fictitious sites, and keep “sample data” markers everywhere so they cannot pass as real audits.

The guides teach transferable methods — how to check, judge and verify — using public, stable behavior of search engines and AI assistants. They are not a substitute for an audit of your specific site, and they cannot answer whether a specific AI system will cite you next week.

• Assistants personalize and change behavior without notice; a snapshot describes that day, not the system.
• Answer generation is non-deterministic: the same prompt can yield different answers and citations.
• Regional and language differences are large; one market’s sample says little about another.
• Mention, citation and accuracy are different questions and are reported separately.

The roadmap describes what we plan to build under these principles. The waitlist is how you hear when it ships.